Cultivation of Chlorella vulgaris in anaerobically digested gelatin industry wastewater

This work aimed to study the effect of using anaerobically digested gelatin industry wastewater as a culture medium for the cultivation of Chlorella vulgaris microalgae in bubble column photobioreactors (PBRs). Batch experiments were carried out to determine the growth kinetics by inoculating microalgae in wastewater prepared with different dilutions and supplemented with Bold’s Basal Medium (BBM). From the values of the saturation constants (KS1⁄4 50.25 mgN-NH4·L ) and substrate inhibition (KI1⁄4 28.12 mgNNH4 þ·L ) obtained in the adjustment to the Andrews kinetic model (R1⁄4 0.817), the PBRs achieved specific maximum growth rates (μmax) of 0.343 d , biomass productivity of 0.141 g·L ·d , lipid content of 12.45%, lipid productivity of 17.63 mg·L ·d 1 and instantaneous ammoniacal nitrogen consumption rates of 20.06 and 14.22 mg·L ·d . The addition of wastewater to the culture medium provided an increase in biomass productivity of 57.45% in relation to the negative control. The results obtained demonstrate the high efficiency of C. vulgaris in the removal of nitrogenous compounds and the potential of using anaerobically digested gelatin industry wastewater in the production of microalgae biomass.

After filtration of the wastewater, the sterilization was carried out by adding 1.0 mL·L À1 of sodium hypochlorite (5%). The solution was manually homogenized with a glass rod and allowed to stand for 12 hours for complete elimination of microorganisms (species control). A higher concentration of sodium hypochlorite than conventional sterilization methods was used to promote irreversible oxidation of sulphides. After this period, an indicator test was performed to verify the absence of chlorine in the solution.
In cases where residual chlorine was observed, the solution was neutralized with sodium thiosulfate in order to prevent possible interference in the growth of microalgae.
Finally, the wastewater was supplemented with modified Bold's Basal Medium (BBM) to optimize the growth of microorganisms and to supply the nutritional deficiencies of the effluent. Table 1

Calibration curve
The microalgae growth was monitored based on a calibration curve (R 2 ¼ 0.9996) designed to convert the

Inoculum adaptation
Before the kinetic test, the inoculum was adapted to the wastewater. The adaptation process was performed by sequential batches in a 1.0 L bubble column photobioreactor (PBR). The PBR used was fed with increasing concentrations of ammoniacal nitrogen (C 1 ¼ 7.05 mg·L À1 , and kept in operation until the exponential growth phase and an approximate biomass concentration of 1.2 g·L À1 in each condition was reached. Then, the batches were discontinued, and the biomass was left for one hour to settle to the bottom of the reactor. After sedimentation of the biomass,

Microbial growth kinetics assays
The objective of this assay was to evaluate microalgae growth by monitoring the cumulative biomass production (X ) and to determine the optimal substrate concentration (S opt ). For the kinetic assay, six different concentrations of ammoniacal (N-NH x ) nitrogen were studied: C 1 ¼ 7.05 mgN-NH 4 þ ·L À1 , 169.14 mgN-NH 4 þ ·L À1 . Nutrient solutions composed of Bold's Basal Medium (BBM) were prepared as described in Table 1  feed, and instead, pneumatic agitation was performed.
Thus, the only carbon source for the microorganisms was the organic carbon contained in the wastewater, sodium bicarbonate of the basal medium, and the CO 2 present in the atmosphere introduced by the headspace, due to the constant shaking of the bottles (complete mixture). The incubation lasted seven days until biomass production ceased.

Bubble column photobioreactors (PBRs)
The microalgae were grown for the dispersion of air microbubbles in the system. Figure 1 shows the schematic drawing of the bubble column PBRs used.

Kinetic modelling of microbial growth
The biomass production profiles of each condition were adjusted to the modified Gompertz kinetic model (Equation (1)) described by Mansouri (), using Origin 8.0 software: In this expression, Y is the concentration of biomass at time t (g·L À1 ); C is the asymptote of ln(X/X o ) at time t; μ max is the maximum specific growth rate (d À1 ); λ is the latency phase duration time (d); t is the batch reaction time (d); and exp is the Euler number.
Determining the maximum specific growth rates (μ max ), a graph was constructed that relates the initial substrate concentrations (S) to the maximum specific growth rates (μ max ) of each condition, and the adjustments to known kinetic models were tested by obtaining kinetic parameters and optimal substrate concentration (S opt ). As substrate inhibition was observed, the adjusted kinetic model used was the Andrews model (Equation (2)), and the S opt calculated by Equation (3). The adjustment to the Andrews model was performed in the Origin 8.0 software.
In these expressions, μ is the specific growth rate (d À1 ); K s is the saturation constant defined by the Monod equation ; and S opt is the optimal substrate concentration for the operation of bubble column photobioreactors (mgN-NH 4 þ ·L À1 ).
To estimate the parameters of the models, the Levenberg-Marquardt (LM) algorithm was used to adjust the expressions of the equations to the experimental results.
The accuracy of the predictive models was investigated using the Adjusted R-Square (R 2 ) and the Standard Error (SE).

Kinetic modelling of substrate consumption
The consumption of ammoniacal nitrogen (N-NH x ) and nitrate (N-NO 3 À ) were adjusted to the first-order kinetic expression and to the modified first-order kinetic expression, presented in Equation (4) and (5) respectively. Both adjustments were performed using the Origin 8.0 software.
In this expression, S is the substrate concentration (mg·L À1 ); S o is the initial substrate concentration (mg·L À1 ); S R is the residual substrate concentration (mg·L À1 ); t is

time (d)
; and k is the decay rate constant (d À1 ). From the k results obtained in the kinetic modelling, the instantaneous substrate consumption rates (r S ) were calculated using Equation (6): After separating the phases, excess water was removed from the surface, and then, the decanted biomass was filtered using a filtration system composed of a Kitasato flask and a Büchner funnel with qualitative filter paper (Unifil ETQ -185 mm).
The wet biomass was uniformly distributed over a watch glass and taken to dry in an oven at a temperature of 55-60 C for a period of 24 hours and subsequently placed in desiccators in order to stabilize the mass at room temperature (Zorn et al. ). After reaching constant mass, the biomass was crushed using a porcelain and pistil gral until it reached a homogeneous powder.

Lipids extraction
The method used in the extraction of lipids via organic solvents was developed following the methodology by Zorn . This methodology promotes a higher lipid yield during the extraction process and is characterized by the use of microalgal biomass with 64% humidity, a chloroform:methanol:water ratio of 5.7:3:1, a total of 33 mL of solvents, and only 70 minutes of ultrasound.

Biomass and lipid productivity
The biomass concentration was quantified by spectrophotometry and the lipid content by extraction with organic solvents. These values were leveraged for calculation of the biomass productivity (P B ) through Equation (7) and the lipid productivity (P L ) through Equation (8). The calculation of the lipid productivity was performed by multiplying the biomass productivity (P B ) by the mass fraction of lipids (F M ). The calculation of the mass fraction of lipids (F M ) was done using Equation (9): In these expressions, P B is the biomass productivity (g·L À1 ·d À1 ); P L is the lipids productivity (g·L À1 ·d À1 ); X F is the final concentration of biomass (g·L À1 ); X O is the initial concentration of biomass (g·L À1 ); F M is the mass fraction of lipids present in the biomass (%); L E is the mass of lipids extracted (g); X E is the amount of biomass used in the extraction (g); and t is the time (d).

Statistical analysis
To statistically confirm which conditions had the greatest influence on the biomass production process (response variable), ANOVA (analysis of variance) was used. The ANOVA was performed using Minitab software version 17 and a 95% confidence interval (p-value ¼ 0.05). The biomass productivity (P B ) data, used in the statistical analysis, were calculated (Equation (7)) from the biomass concentration results obtained by spectrophotometry during the monitoring of the cultures in the bubble column photobioreactors (PBRs).

Microbial growth kinetics
The biomass production results obtained in each condition of the kinetic assay were adjusted to the modified Gompertz model to obtain the maximum specific growth rates (μ max ).
Each condition tested (C 1 , C 2 , C 3 , C 4 , C 5 , and C 6 ) has different concentrations of ammoniacal nitrogen (N-NH x ), as described in the section above on 'Microbial growth kinetics assays'. Considering that cell death occurred in condition C 6 , the experimental data for microbial growth were not obtained, and therefore, no adjustment was made. The nonlinear adjustments performed for each condition of the kinetic assay are shown in Figure 2. Table 2 was then created for better visualization of the maximum specific growth rate (μ max ) and latency phase duration time (λ) results, obtained from Equation (1)  Analyzing the results of the latency phase duration time (λ), it can be observed that the time was relatively low under all conditions, proving that the inoculum adaptation process to the wastewater was efficient. It is noted that the control showed the largest latency phase among the tested conditions and that as the availability of ammoniacal nitrogen in the medium increased, the latency phase (λ) was reduced. This is likely due to the fact that microalgae have higher affinity to ammoniacal nitrogen compared with oxidized nitrogen compounds, which require a higher energy consumption in the assimilation process (Maestrini et al.

).
However, conditions C 4 and C 5 presented considerably higher ammoniacal nitrogen concentrations and demonstrated a gradual increase in the latency phase.
Thus, after evaluating the behavior of maximum specific growth rates (μ max ) and latency phase duration time (λ), the Andrews model (Equation (2)) was proposed and adjusted to describe the behavior of microorganisms. From the model, the kinetic parameters μ max , K s , and K i were determined to be 1.27 d À1 , 50.25 mg·L À1 , and 28.12 mg·L À1 , respectively. The fit to the Andrews model is illustrated in Figure 3.
Tam & Wong () also performed experiments to evaluate the influence of ammoniacal nitrogen concentration on the 169.14 ± 1.87 0.000 ± 0.000 0.00 ± 0.00 0.0000 The dissolved oxygen (DO) of the reactor flasks was also monitored daily during the kinetic test cultures, presenting no changes in relation to their initial concentration (DO ¼ 5.69 mgO 2 ·L À1 ). Thus, it is possible to state that the proposed headspace degassing strategy was efficient.

Cultivation in bubble column photobioreactors (PBRs)
Kinetic modelling of microalgae growth

photobioreactors (PBRs) fed with wastewater and with Bold's
Basal Medium (control) are illustrated in Figure 4.
From the nonlinear adjustments of the kinetic models, the maximum specific growth rates (μ max ) and the duration of the latency phase (λ), obtained in the preparation of Table 3, were obtained for better visualization and interpretation of the results. In addition to the kinetic parameters obtained during kinetic modelling, the biomass productivity (P B ) of the photobioreactors was also calculated and is presented in Table 3 Comparing the results of μ max and X F obtained, it is possible to verify that in the microalgae cultivations in bubble column photobioreactors (PBRs), the addition of ammoniacal nitrogen (N-NH x ) provided an increase in both the maximum specific growth rate (0.34 d À1 ) and the production of biomass (1.83 g·L À1 ) with respect to the negative control. This is attributed to the greater affinity of microorganisms to nitrogen in reduced forms (N-NH x ), which lessens the energy expended to reduce and assimilate oxidized nitrogen compounds (N-NO 2 À and N-NO 3 À ). Therefore, although it is necessary to dilute the effluent, the increase in the maximum specific growth rates (μ max ) and the production of biomass (X F ) demonstrate the potential of cultivating C. vulgaris in anaerobically digested wastewater from food industries.
Mansouri () modelled the growth of photoautotrophically grown C. vulgaris microalgae in 20 L airlift photobioreactors, and they used an artificial culture medium (BG-11) with only nitrate (1.5 g·L À1 ) as a nitrogen source. In this work, the kinetic model that provided the best fit was the Baranyi model (R 2 ¼ 0.989); however, the and as two decay phases were observed during the consumption of ammoniacal nitrogen, two modified first-order kinetic expressions (Equation (5)) in series were also adjusted.
The first decay phase was from day 1 to day 5, where microalgae still assimilated nitrate, and the second phase of decay was from day 5 to 10, where preference was given to nitrogen in the reduced form (N-NH x ). The second phase demonstrated the interruption of nitrate assimilation, coincidentally resulting in lack of adjustment of the model during this period. The preference for ammoniacal nitrogen began at day 5 due to the greater availability of nitrate in the medium as compared with ammoniacal nitrogen in the beginning of the batch, which favored the mass transfer between substrate and microorganism. Thus, two modified first-order kinetic expression adjustments were proposed for the phases of N-NH x consumption, named as Phase 1 and Phase 2 (before and after the interruption of nitrate assimilation), respectively. Based on the kinetic parameters obtained, Table 4 is provided to better visualize these results. The analysis of the errors of the kinetic parameters was performed using Standard Error (SE) and the Adjusted R-Square (R 2 ).
Analyzing the Adjusted R-Square (R 2 ), it is possible to verify that the modified first-order kinetic expression adjusted satisfactorily to the two phases of substrate consumption in the bubble column photobioreactors (PBRs). Although the adjustment to single phase achieved a good Adjusted R-Square (R 2 ), the analysis of substrate consumption by parts proved to be more adequate to explain the consumption of nitrogen forms present in the bulk liquid.
Regarding the ammoniacal nitrogen decay rate constants (k), it was observed that from the fifth day on when the nitrate consumption was interrupted (Figure 6), the microalgae channelled all their energy into the assimilation of ammoniacal nitrogen, and consequently, there was an increase in the decay rate constant from 0.453 to 0.616 d À1 . However, the initial substrate concentration (C o ) was lower in phase 2, and although the k values increased, the instantaneous substrate consumption rates (r S ) decreased. It is important to note, however, that both r S results were considerably superior to those obtained by Choi & Lee () in their study.
Interpreting the behavior of the substrate decay and cell growth curves shown in Figure 6, it is possible to observe that during the first five days, the microalgae consumed nitrate and ammoniacal nitrogen. During the period between days 3 and 5, the consumption of both ammoniacal   In this study, the accumulated lipid content was 16.4%, and the lipid productivity was 0.0126 g·L À1 ·d À1 .
The lipid content was higher than that obtained in the present project due to the low availability of nitrogen in the domestic effluent; however, the lipid productivity was lower because although the lipid content percentage was high, the biomass productivity was low (0.0552 g·L À1 ·d À1 ).
This phenomenon is explained by the fact that the availability of substrate is a limiting step in the balance of growth and production of fatty acids, and thus, the biomass and fatty acid synthesis pathways compete for the same substrates (Tan & Lee ).
In the study by Guimarães et al. (), biomass production ended, and the nutrient availability was reduced while the photobioreactors were kept in operation for a few more days in order to promote the stress of microorganisms. The difference from the present project is notable, as the batch was interrupted at the exact moment when the microorganisms reached the stationary growth phase.
Even after the stagnation of biomass production, there was approximately 80 mg·L À1 of nitrate, which prevented the microorganisms from being exposed to critical conditions of low substrate availability in the culture environment.

CONCLUSION
The cultivation of microalgae C. vulgaris in bubble column photobioreactors (PBRs) fed with wastewater nutrients proved to be a viable and efficient alternative for both biomass production and the removal of nitrogenous compounds from water, especially ammoniacal nitrogen (N-NH x ). The Andrews model satisfactorily fitted the biomass growth data by presenting a high Adjusted R-Square (R 2 ), correctly describing the microbial kinetics and providing the optimized process operation. It can be observed that the use of anaerobically digested gelatin industry wastewater resulted in higher specific growth rates and higher biomass productivity, however, the accumulated lipid content per gram of biomass was low despite lipid productivity having been reasonable. Future studies can further elaborate on the potential benefits and limitations of using food industry and process wastewaters for simultaneous nutrient recovery and algal biomass production at larger scale and different cultivation system configurations in order to evaluate system viability, economics, and sustainability.

DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.